Subways, Strikes, and Slowdowns: The Impacts of Public Transit on ...

NBER WORKING PAPER SERIES

SUBWAYS, STRIKES, AND SLOWDOWNS: THE IMPACTS OF PUBLIC TRANSIT ON TRAFFIC CONGESTION

Michael L. Anderson Working Paper 18757

NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 February 2013

I thank Ken Small, Lowell Taylor, Matt Turner, and participants at the 13th Occasional California Workshop on Environmental and Resource Economics, the 2012 AERE Summer Conference, Texas A&M, and the University of Houston for valuable suggestions. Any errors in the paper are the author's. The views expressed herein are those of the author and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peerreviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications. ? 2013 by Michael L. Anderson. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including ? notice, is given to the source.

Subways, Strikes, and Slowdowns: The Impacts of Public Transit on Traffic Congestion Michael L. Anderson NBER Working Paper No. 18757 February 2013 JEL No. R41,R42,R48

ABSTRACT

Public transit accounts for only 1% of U.S. passenger miles traveled but nevertheless attracts strong public support. Using a simple choice model, we predict that transit riders are likely to be individuals who commute along routes with the most severe roadway delays. These individuals' choices thus have very high marginal impacts on congestion. We test this prediction with data from a sudden strike in 2003 by Los Angeles transit workers. Estimating a regression discontinuity design, we find that average highway delay increases 47% when transit service ceases. This effect is consistent with our model's predictions and many times larger than earlier estimates, which have generally concluded that public transit provides minimal congestion relief. We find that the net benefits of transit systems appear to be much larger than previously believed.

Michael L. Anderson Department of Agricultural and Resource Economics 207 Giannini Hall, MC 3310 University of California, Berkeley Berkeley, CA 94720 and NBER mlanderson@berkeley.edu

1. INTRODUCTION It is a stylized fact in the transportation literature that mass transit attracts a

disproportionate share of public funds but carries a negligible fraction of commuters. In 2010, public transit received 23% of federal highway and transit outlays but accounted for 1% of passenger miles traveled (U.S. Department of Transportation 2009, 2011a, 2011b). State, local, and federal subsidies exceed $40 billion per year and cover 63% of operating costs and 100% of capital costs. Even in Washington, DC ? which boasts the second busiest metro system in the United States ? transit accounts for only 5% of passenger miles traveled (American Public Transportation Association 2011; Schrank, Lomax, and Eisele 2011).

Public transit subsidies nevertheless remain popular in many areas. For example, in 2008 67% of Los Angeles County residents voted to allocate $26 billion to transit over 30 years. Why is there such deep public support for transit subsidies if few voters are frequent riders? The simplest explanation is the possibility of congestion relief ? commuters may expect to benefit from reduced congestion even if they rarely use public transit themselves.1 A large body of transportation and economic research, however, concludes that public transit has little effect on reducing congestion, calling into question its heavy subsidy rate (Rubin, Moore, and Lee 1999; Stopher 2004; Small 2005; Winston and Maheshri 2007).

An important detail that has received little attention in the existing literature is that commuters on different roadways in the same metropolitan area face sharply different levels of congestion during peak hours. This paper presents a simple choice model in which commuters face differing levels of congestion and choose either to drive or take transit. Calibrating the model using data from the Los Angeles metro area, we predict effects on congestion that are approximately six times larger than a model that does not account for heterogeneity in congestion levels. This prediction is much larger than previous estimates, and the qualitative conclusion is robust to wide variations in parameter values. The intuition is straightforward: Transit is most attractive to commuters who face the worst congestion, so a disproportionate number of transit riders are commuters who would otherwise have to drive on the most congested roads at the most congested times. Since drivers on heavily congested roads have a much higher marginal impact on congestion than drivers on the average road, transit has a large impact on reducing traffic congestion.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

1 This possibility is perhaps best summarized by the title of a satirical article in the November 29, 2000, issue of The Onion, "Report: 98 Percent of U.S. Commuters Favor Public Transportation for Others." Other factors that may explain local support of capital investment in transit include high federal matching rates, a combination of concentrated economic rents and dispersed costs, and the political appeal of "ribbon cutting" ceremonies (Taylor 2004; Baum-Snow and Kahn 2005; Winston and Maheshri 2007).

!

1

We test our predictions using freeway speed data from a 2003 strike by Los Angeles County Metropolitan Transportation Authority (MTA) workers. In October 2003, MTA workers began a strike that lasted 35 days and shut down MTA bus and rail lines. Using hourly data on traffic speeds for all major Los Angeles freeways, we estimate a regression discontinuity (RD) design using time as the running variable. We find an abrupt increase in average delays of 47% (0.19 minutes per mile) during peak periods. This increase persists through the end of the strike, and the estimate ? consistent with the predictions of our model ? is many times larger than estimates in the existing literature. The effects are largest on freeways that parallel transit lines with heavy ridership, and they are small and statistically insignificant during the same period in neighboring counties unaffected by the transit strike.

Our estimates imply that the total congestion relief benefit of operating the Los Angeles transit system is between $1.2 billion to $4.1 billion per year, or $1.20 to $4.10 per peak-hour transit passenger mile. We consider the potential gap between the short-run effect of ceasing transit provision (i.e., our estimates) and the long-run effect of a permanent shutdown. We find that reducing the long-run effect to less than 50% of the short-run effect's lower bound requires implausibly large elasticities of travel with respect to travel costs. We consider the net benefits of constructing the Los Angeles rail system and conclude ? contrary to the existing literature on rail capital investment ? that they are large and positive.

On a broader scale, our findings demonstrate that in contexts in which policymakers encourage adoption of activities that mitigate negative externalities, considering who adopts the mitigating activity is critical in determining a policy's expected benefits. We close with a brief discussion of other contexts in which selection into mitigating activities may have large impacts on predicted benefits.

2. BACKGROUND Existing economic research on the effects of transit on traffic congestion falls into two

categories: model-based estimates and empirical estimates. Examples of the former include Nelson et al. (2007) and Parry and Small (2009). Parry and Small (2009) develop an analytical model of an urban transportation system and compute the optimal transit operating subsidy. The model takes as inputs average speeds, costs, and price and service elasticities. One input is the effect of transit on relieving traffic congestion. They compute this effect using assumptions about substitution between transportation modes and engineering estimates relating average delays and marginal congestion impacts. In Los Angeles, the congestion relief externality of traveling 1 mile on transit during peak hours is computed at 1.7 person-

!

2

minutes of reduced traffic delays.2 Aggregating this figure across all peak-period transit passengers implies that transit reduces average delay by approximately 5% (0.025 minutes per mile). Nelson et al. explore the potential benefits of the Washington, DC transit system using a simulation model in which travel decisions are modeled as a nested logit tree. The model takes as inputs demand response parameters from the literature, and it is calibrated to match aggregate Washington travel patterns. Applying a relationship between traffic flows and traffic speeds similar to that used by Parry and Small, Nelson et al. calculate that the Washington transit system reduces total congestion by 184,000 person-hours per day, or 2.0 person-minutes per peak transit passenger mile carried. This figure is close to the figure implied by Parry and Small.

Researchers employing empirical approaches include Winston and Langer (2006) and Duranton and Turner (2011). Using metropolitan-area data, these authors regress total congestion or vehicle miles traveled (VMT) on measures of transit capacity. They reach varying conclusions. Winston and Langer estimate that rail lines reduce congestion but that bus lines increase congestion. The net effect of transit systems is thus to increase congestion. Duranton and Turner focus on testing the "fundamental law of road congestion" ? the hypothesis that the primary determinant of VMT in most cities is roadway capacity. They also estimate a positive relationship between bus fleet size and VMT. To address the potential endogeneity of transit provision, they instrument for bus fleets using an area's 1972 Democratic vote share. The relationship between bus fleets and VMT then becomes statistically insignificant and of variable sign, though the instrument is not powerful enough to rule out the possibility that 1 passenger mile traveled on transit removes substantially more than 1 VMT from roadways.3 These findings are nevertheless consistent with our own findings, which suggest that transit has a minimal impact on total VMT but a large impact on

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

2 In Los Angeles, Parry and Small assume that each passenger mile traveled on transit diverts approximately 0.9 passenger miles from roadways. The average peak period delay in Los Angeles is 0.5 minutes per mile, and estimates from the literature relating traffic flows and traffic speeds suggest that the marginal effect on total delay of adding an additional vehicle to the road is 3.7 times the average delay. The congestion relief externality of traveling 1 mile on transit during peak hours is thus (?0.9 auto passenger miles/transit passenger mile * 0.5 mins avg delay/passenger mile * 3.7 mins increased delay/min avg delay) = ?1.7 mins increased delay/transit passenger mile. 3 Duranton and Turner estimate that increasing a city's bus fleet by 18 buses (10%) changes annual VMT by ? 1.3% to 0.6% (?35 million to 16 million VMT) in their most precise instrumental variables (IV) specification and ?0.7% to 4.9% (?18 million to 133 million VMT) in their least precise IV specification (these figures correspond to 95% confidence intervals). An average city bus carries 0.3 million passenger miles per year (American Public Transit Association 2011), so adding 5.4 million passenger miles of bus travel could decrease VMT by up to 35 million miles in the most precise specification and up to 18 million miles in the least precise specification.

!

3

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download